Hyperspectral full-polarization image compression and reconstruction method for optimizing sparse base through machine learning

Abstract

The invention discloses a hyperspectral full-polarization image compression and reconstruction method for optimizing a sparse base through machine learning. The method comprises the steps of combininga quarter-wave plate and a device with a linear polarization characteristic to image an image on a detector, and realizing different full-polarization modulation modes by switching a fast axis angleof the quarter-wave plate and / or a light transmission axis angle of the device with the linear polarization characteristic; processing the full-polarization local image of any waveband by adopting thefull-polarization modulation mode to obtain compression information; and optimizing the sparse base by adopting a particle swarm algorithm, wherein the optimized sparse base enables the full-polarization local image reconstructed by utilizing the compressed information to approach the original image of the full-polarization local image. During application, polarization modulation is carried out on a hyperspectral full-polarization image by adopting the full-polarization modulation mode to obtain compression information, and the reconstructed hyperspectral full-polarization image is obtained by utilizing the optimized sparse basis. By adopting the method, the reconstruction of four Stokes parameters of the hyperspectral image can be realized, and the reconstruction precision of the four Stokes parameters is improved.

Description

technical field [0001] The invention belongs to the technical field of hyperspectral polarization imaging, and in particular relates to a method for compressing and reconstructing hyperspectral full-polarization images based on machine learning to optimize sparse bases, so as to realize the reconstruction of full Stokes parameters of hyperspectral images. Background technique [0002] Hyperspectral full-polarization imaging technology is an advanced technology used to obtain four-dimensional data of the target scene, which includes two-dimensional spatial information, one-dimensional spectral information and polarization information represented by four Stokes parameters. Compared with traditional hyperspectral imaging methods, the additional polarization information can characterize surface roughness, electrical conductivity, molecular distribution and material composition, etc. Because hyperspectral full-polarization imaging technology has the advantage of identifying subtl...

AI Technical Summary

Problems solved by technology

However, the measurement of circular polarization information greatly increases the complexity of the system and the loss of light energy, and in the process of reconstructing the four Stokes parameters, the sampling time is long and the information is redundant due to the limitation of sparse basis selection

Images